Construction of hybrid models based on cascade technique using basic machine learning models: An application as photocurrent density predictor of the photoelectrode in PEC cell

• Published in: Materials today communications Vol. 41; p. 110643
• Main Authors: Sahu, Nepal, Azad, Chandrashekhar, Kumar, Uday
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.12.2024
• Subjects: Cascade technique; Hybrid model; Machine learning; Photocurrent density prediction; Photoelectrochemical water splitting; Cascade technique; Hybrid model; Photocurrent density prediction; Photoelectrochemical water splitting; Machine learning
• ISSN: 2352-4928; 2352-4928
• DOI: 10.1016/j.mtcomm.2024.110643

Photocurrent density is the performance assessing metric of a Photoelectrochemical (PEC) water splitting system. For the design and development of an efficient PEC system, it is essential to achieve high photocurrent density of a photoelectrode. Machine Learning (ML) can play a significant role in determining the performance of the photoelectrode through predicting the photocurrent density. Here we have developed three binary hybrid regression models by cascading two single models out of the four basic single ML models (MLP, Adaboost, Ridge, and Elastic Net) for the prediction of the photocurrent density. The performance of the three hybrid models is better than each individual model. The hybrid model of Adaboost and MLP is outperformed among all other hybrid models with R2, MAE and RMSE equal to 96.86%, 0.3319 and 0.5542 respectively which is higher than the latest reported of R2 equal to 95%. Further, the best hybrid model and its constituent models (Adaboost and MLP) were subjected to SHAP analysis and the feature effectiveness is discussed. [Display omitted]